cpython/Lib/asyncio/base_subprocess.py

285 lines
8.6 KiB
Python
Raw Normal View History

2013-10-30 18:56:49 -03:00
import collections
import subprocess
import warnings
2013-10-30 18:56:49 -03:00
from . import protocols
from . import transports
from .log import logger
2013-10-30 18:56:49 -03:00
class BaseSubprocessTransport(transports.SubprocessTransport):
def __init__(self, loop, protocol, args, shell,
stdin, stdout, stderr, bufsize,
asyncio: sync with Tulip Issue #23347: send_signal(), kill() and terminate() methods of BaseSubprocessTransport now check if the transport was closed and if the process exited. Issue #23347: Refactor creation of subprocess transports. Changes on BaseSubprocessTransport: * Add a wait() method to wait until the child process exit * The constructor now accepts an optional waiter parameter. The _post_init() coroutine must not be called explicitly anymore. It makes subprocess transports closer to other transports, and it gives more freedom if we want later to change completly how subprocess transports are created. * close() now kills the process instead of kindly terminate it: the child process may ignore SIGTERM and continue to run. Call explicitly terminate() and wait() if you want to kindly terminate the child process. * close() now logs a warning in debug mode if the process is still running and needs to be killed * _make_subprocess_transport() is now fully asynchronous again: if the creation of the transport failed, wait asynchronously for the process eixt. Before the wait was synchronous. This change requires close() to *kill*, and not terminate, the child process. * Remove the _kill_wait() method, replaced with a more agressive close() method. It fixes _make_subprocess_transport() on error. BaseSubprocessTransport.close() calls the close() method of pipe transports, whereas _kill_wait() closed directly pipes of the subprocess.Popen object without unregistering file descriptors from the selector (which caused severe bugs). These changes simplifies the code of subprocess.py.
2015-01-29 19:05:19 -04:00
waiter=None, extra=None, **kwargs):
2013-10-30 18:56:49 -03:00
super().__init__(extra)
self._closed = False
2013-10-30 18:56:49 -03:00
self._protocol = protocol
self._loop = loop
asyncio: sync with Tulip Issue #23347: send_signal(), kill() and terminate() methods of BaseSubprocessTransport now check if the transport was closed and if the process exited. Issue #23347: Refactor creation of subprocess transports. Changes on BaseSubprocessTransport: * Add a wait() method to wait until the child process exit * The constructor now accepts an optional waiter parameter. The _post_init() coroutine must not be called explicitly anymore. It makes subprocess transports closer to other transports, and it gives more freedom if we want later to change completly how subprocess transports are created. * close() now kills the process instead of kindly terminate it: the child process may ignore SIGTERM and continue to run. Call explicitly terminate() and wait() if you want to kindly terminate the child process. * close() now logs a warning in debug mode if the process is still running and needs to be killed * _make_subprocess_transport() is now fully asynchronous again: if the creation of the transport failed, wait asynchronously for the process eixt. Before the wait was synchronous. This change requires close() to *kill*, and not terminate, the child process. * Remove the _kill_wait() method, replaced with a more agressive close() method. It fixes _make_subprocess_transport() on error. BaseSubprocessTransport.close() calls the close() method of pipe transports, whereas _kill_wait() closed directly pipes of the subprocess.Popen object without unregistering file descriptors from the selector (which caused severe bugs). These changes simplifies the code of subprocess.py.
2015-01-29 19:05:19 -04:00
self._proc = None
self._pid = None
asyncio: sync with Tulip Issue #23347: send_signal(), kill() and terminate() methods of BaseSubprocessTransport now check if the transport was closed and if the process exited. Issue #23347: Refactor creation of subprocess transports. Changes on BaseSubprocessTransport: * Add a wait() method to wait until the child process exit * The constructor now accepts an optional waiter parameter. The _post_init() coroutine must not be called explicitly anymore. It makes subprocess transports closer to other transports, and it gives more freedom if we want later to change completly how subprocess transports are created. * close() now kills the process instead of kindly terminate it: the child process may ignore SIGTERM and continue to run. Call explicitly terminate() and wait() if you want to kindly terminate the child process. * close() now logs a warning in debug mode if the process is still running and needs to be killed * _make_subprocess_transport() is now fully asynchronous again: if the creation of the transport failed, wait asynchronously for the process eixt. Before the wait was synchronous. This change requires close() to *kill*, and not terminate, the child process. * Remove the _kill_wait() method, replaced with a more agressive close() method. It fixes _make_subprocess_transport() on error. BaseSubprocessTransport.close() calls the close() method of pipe transports, whereas _kill_wait() closed directly pipes of the subprocess.Popen object without unregistering file descriptors from the selector (which caused severe bugs). These changes simplifies the code of subprocess.py.
2015-01-29 19:05:19 -04:00
self._returncode = None
self._exit_waiters = []
self._pending_calls = collections.deque()
2013-10-30 18:56:49 -03:00
self._pipes = {}
asyncio: sync with Tulip Issue #23347: send_signal(), kill() and terminate() methods of BaseSubprocessTransport now check if the transport was closed and if the process exited. Issue #23347: Refactor creation of subprocess transports. Changes on BaseSubprocessTransport: * Add a wait() method to wait until the child process exit * The constructor now accepts an optional waiter parameter. The _post_init() coroutine must not be called explicitly anymore. It makes subprocess transports closer to other transports, and it gives more freedom if we want later to change completly how subprocess transports are created. * close() now kills the process instead of kindly terminate it: the child process may ignore SIGTERM and continue to run. Call explicitly terminate() and wait() if you want to kindly terminate the child process. * close() now logs a warning in debug mode if the process is still running and needs to be killed * _make_subprocess_transport() is now fully asynchronous again: if the creation of the transport failed, wait asynchronously for the process eixt. Before the wait was synchronous. This change requires close() to *kill*, and not terminate, the child process. * Remove the _kill_wait() method, replaced with a more agressive close() method. It fixes _make_subprocess_transport() on error. BaseSubprocessTransport.close() calls the close() method of pipe transports, whereas _kill_wait() closed directly pipes of the subprocess.Popen object without unregistering file descriptors from the selector (which caused severe bugs). These changes simplifies the code of subprocess.py.
2015-01-29 19:05:19 -04:00
self._finished = False
2013-10-30 18:56:49 -03:00
if stdin == subprocess.PIPE:
self._pipes[0] = None
2013-10-30 18:56:49 -03:00
if stdout == subprocess.PIPE:
self._pipes[1] = None
2013-10-30 18:56:49 -03:00
if stderr == subprocess.PIPE:
self._pipes[2] = None
asyncio: sync with Tulip Issue #23347: send_signal(), kill() and terminate() methods of BaseSubprocessTransport now check if the transport was closed and if the process exited. Issue #23347: Refactor creation of subprocess transports. Changes on BaseSubprocessTransport: * Add a wait() method to wait until the child process exit * The constructor now accepts an optional waiter parameter. The _post_init() coroutine must not be called explicitly anymore. It makes subprocess transports closer to other transports, and it gives more freedom if we want later to change completly how subprocess transports are created. * close() now kills the process instead of kindly terminate it: the child process may ignore SIGTERM and continue to run. Call explicitly terminate() and wait() if you want to kindly terminate the child process. * close() now logs a warning in debug mode if the process is still running and needs to be killed * _make_subprocess_transport() is now fully asynchronous again: if the creation of the transport failed, wait asynchronously for the process eixt. Before the wait was synchronous. This change requires close() to *kill*, and not terminate, the child process. * Remove the _kill_wait() method, replaced with a more agressive close() method. It fixes _make_subprocess_transport() on error. BaseSubprocessTransport.close() calls the close() method of pipe transports, whereas _kill_wait() closed directly pipes of the subprocess.Popen object without unregistering file descriptors from the selector (which caused severe bugs). These changes simplifies the code of subprocess.py.
2015-01-29 19:05:19 -04:00
# Create the child process: set the _proc attribute
try:
self._start(args=args, shell=shell, stdin=stdin, stdout=stdout,
stderr=stderr, bufsize=bufsize, **kwargs)
except:
self.close()
raise
self._pid = self._proc.pid
2013-10-30 18:56:49 -03:00
self._extra['subprocess'] = self._proc
asyncio: sync with Tulip Issue #23347: send_signal(), kill() and terminate() methods of BaseSubprocessTransport now check if the transport was closed and if the process exited. Issue #23347: Refactor creation of subprocess transports. Changes on BaseSubprocessTransport: * Add a wait() method to wait until the child process exit * The constructor now accepts an optional waiter parameter. The _post_init() coroutine must not be called explicitly anymore. It makes subprocess transports closer to other transports, and it gives more freedom if we want later to change completly how subprocess transports are created. * close() now kills the process instead of kindly terminate it: the child process may ignore SIGTERM and continue to run. Call explicitly terminate() and wait() if you want to kindly terminate the child process. * close() now logs a warning in debug mode if the process is still running and needs to be killed * _make_subprocess_transport() is now fully asynchronous again: if the creation of the transport failed, wait asynchronously for the process eixt. Before the wait was synchronous. This change requires close() to *kill*, and not terminate, the child process. * Remove the _kill_wait() method, replaced with a more agressive close() method. It fixes _make_subprocess_transport() on error. BaseSubprocessTransport.close() calls the close() method of pipe transports, whereas _kill_wait() closed directly pipes of the subprocess.Popen object without unregistering file descriptors from the selector (which caused severe bugs). These changes simplifies the code of subprocess.py.
2015-01-29 19:05:19 -04:00
if self._loop.get_debug():
if isinstance(args, (bytes, str)):
program = args
else:
program = args[0]
logger.debug('process %r created: pid %s',
program, self._pid)
asyncio: sync with Tulip Issue #23347: send_signal(), kill() and terminate() methods of BaseSubprocessTransport now check if the transport was closed and if the process exited. Issue #23347: Refactor creation of subprocess transports. Changes on BaseSubprocessTransport: * Add a wait() method to wait until the child process exit * The constructor now accepts an optional waiter parameter. The _post_init() coroutine must not be called explicitly anymore. It makes subprocess transports closer to other transports, and it gives more freedom if we want later to change completly how subprocess transports are created. * close() now kills the process instead of kindly terminate it: the child process may ignore SIGTERM and continue to run. Call explicitly terminate() and wait() if you want to kindly terminate the child process. * close() now logs a warning in debug mode if the process is still running and needs to be killed * _make_subprocess_transport() is now fully asynchronous again: if the creation of the transport failed, wait asynchronously for the process eixt. Before the wait was synchronous. This change requires close() to *kill*, and not terminate, the child process. * Remove the _kill_wait() method, replaced with a more agressive close() method. It fixes _make_subprocess_transport() on error. BaseSubprocessTransport.close() calls the close() method of pipe transports, whereas _kill_wait() closed directly pipes of the subprocess.Popen object without unregistering file descriptors from the selector (which caused severe bugs). These changes simplifies the code of subprocess.py.
2015-01-29 19:05:19 -04:00
self._loop.create_task(self._connect_pipes(waiter))
def __repr__(self):
info = [self.__class__.__name__]
if self._closed:
info.append('closed')
if self._pid is not None:
info.append(f'pid={self.pid}')
if self._returncode is not None:
info.append(f'returncode={self._returncode}')
elif self._pid is not None:
info.append('running')
else:
info.append('not started')
stdin = self._pipes.get(0)
if stdin is not None:
info.append(f'stdin={stdin.pipe}')
stdout = self._pipes.get(1)
stderr = self._pipes.get(2)
if stdout is not None and stderr is stdout:
info.append(f'stdout=stderr={stdout.pipe}')
else:
if stdout is not None:
info.append(f'stdout={stdout.pipe}')
if stderr is not None:
info.append(f'stderr={stderr.pipe}')
return '<{}>'.format(' '.join(info))
2013-10-30 18:56:49 -03:00
def _start(self, args, shell, stdin, stdout, stderr, bufsize, **kwargs):
raise NotImplementedError
def set_protocol(self, protocol):
self._protocol = protocol
def get_protocol(self):
return self._protocol
def is_closing(self):
return self._closed
2013-10-30 18:56:49 -03:00
def close(self):
if self._closed:
return
self._closed = True
asyncio: sync with Tulip Issue #23347: send_signal(), kill() and terminate() methods of BaseSubprocessTransport now check if the transport was closed and if the process exited. Issue #23347: Refactor creation of subprocess transports. Changes on BaseSubprocessTransport: * Add a wait() method to wait until the child process exit * The constructor now accepts an optional waiter parameter. The _post_init() coroutine must not be called explicitly anymore. It makes subprocess transports closer to other transports, and it gives more freedom if we want later to change completly how subprocess transports are created. * close() now kills the process instead of kindly terminate it: the child process may ignore SIGTERM and continue to run. Call explicitly terminate() and wait() if you want to kindly terminate the child process. * close() now logs a warning in debug mode if the process is still running and needs to be killed * _make_subprocess_transport() is now fully asynchronous again: if the creation of the transport failed, wait asynchronously for the process eixt. Before the wait was synchronous. This change requires close() to *kill*, and not terminate, the child process. * Remove the _kill_wait() method, replaced with a more agressive close() method. It fixes _make_subprocess_transport() on error. BaseSubprocessTransport.close() calls the close() method of pipe transports, whereas _kill_wait() closed directly pipes of the subprocess.Popen object without unregistering file descriptors from the selector (which caused severe bugs). These changes simplifies the code of subprocess.py.
2015-01-29 19:05:19 -04:00
2013-10-30 18:56:49 -03:00
for proto in self._pipes.values():
if proto is None:
continue
2013-10-30 18:56:49 -03:00
proto.pipe.close()
asyncio: sync with Tulip Issue #23347: send_signal(), kill() and terminate() methods of BaseSubprocessTransport now check if the transport was closed and if the process exited. Issue #23347: Refactor creation of subprocess transports. Changes on BaseSubprocessTransport: * Add a wait() method to wait until the child process exit * The constructor now accepts an optional waiter parameter. The _post_init() coroutine must not be called explicitly anymore. It makes subprocess transports closer to other transports, and it gives more freedom if we want later to change completly how subprocess transports are created. * close() now kills the process instead of kindly terminate it: the child process may ignore SIGTERM and continue to run. Call explicitly terminate() and wait() if you want to kindly terminate the child process. * close() now logs a warning in debug mode if the process is still running and needs to be killed * _make_subprocess_transport() is now fully asynchronous again: if the creation of the transport failed, wait asynchronously for the process eixt. Before the wait was synchronous. This change requires close() to *kill*, and not terminate, the child process. * Remove the _kill_wait() method, replaced with a more agressive close() method. It fixes _make_subprocess_transport() on error. BaseSubprocessTransport.close() calls the close() method of pipe transports, whereas _kill_wait() closed directly pipes of the subprocess.Popen object without unregistering file descriptors from the selector (which caused severe bugs). These changes simplifies the code of subprocess.py.
2015-01-29 19:05:19 -04:00
if (self._proc is not None and
# has the child process finished?
self._returncode is None and
# the child process has finished, but the
# transport hasn't been notified yet?
self._proc.poll() is None):
asyncio: sync with Tulip Issue #23347: send_signal(), kill() and terminate() methods of BaseSubprocessTransport now check if the transport was closed and if the process exited. Issue #23347: Refactor creation of subprocess transports. Changes on BaseSubprocessTransport: * Add a wait() method to wait until the child process exit * The constructor now accepts an optional waiter parameter. The _post_init() coroutine must not be called explicitly anymore. It makes subprocess transports closer to other transports, and it gives more freedom if we want later to change completly how subprocess transports are created. * close() now kills the process instead of kindly terminate it: the child process may ignore SIGTERM and continue to run. Call explicitly terminate() and wait() if you want to kindly terminate the child process. * close() now logs a warning in debug mode if the process is still running and needs to be killed * _make_subprocess_transport() is now fully asynchronous again: if the creation of the transport failed, wait asynchronously for the process eixt. Before the wait was synchronous. This change requires close() to *kill*, and not terminate, the child process. * Remove the _kill_wait() method, replaced with a more agressive close() method. It fixes _make_subprocess_transport() on error. BaseSubprocessTransport.close() calls the close() method of pipe transports, whereas _kill_wait() closed directly pipes of the subprocess.Popen object without unregistering file descriptors from the selector (which caused severe bugs). These changes simplifies the code of subprocess.py.
2015-01-29 19:05:19 -04:00
if self._loop.get_debug():
logger.warning('Close running child process: kill %r', self)
try:
self._proc.kill()
except ProcessLookupError:
pass
# Don't clear the _proc reference yet: _post_init() may still run
2013-10-30 18:56:49 -03:00
def __del__(self):
if not self._closed:
warnings.warn(f"unclosed transport {self!r}", ResourceWarning,
source=self)
self.close()
2013-10-30 18:56:49 -03:00
def get_pid(self):
return self._pid
2013-10-30 18:56:49 -03:00
def get_returncode(self):
return self._returncode
def get_pipe_transport(self, fd):
if fd in self._pipes:
return self._pipes[fd].pipe
else:
return None
asyncio: sync with Tulip Issue #23347: send_signal(), kill() and terminate() methods of BaseSubprocessTransport now check if the transport was closed and if the process exited. Issue #23347: Refactor creation of subprocess transports. Changes on BaseSubprocessTransport: * Add a wait() method to wait until the child process exit * The constructor now accepts an optional waiter parameter. The _post_init() coroutine must not be called explicitly anymore. It makes subprocess transports closer to other transports, and it gives more freedom if we want later to change completly how subprocess transports are created. * close() now kills the process instead of kindly terminate it: the child process may ignore SIGTERM and continue to run. Call explicitly terminate() and wait() if you want to kindly terminate the child process. * close() now logs a warning in debug mode if the process is still running and needs to be killed * _make_subprocess_transport() is now fully asynchronous again: if the creation of the transport failed, wait asynchronously for the process eixt. Before the wait was synchronous. This change requires close() to *kill*, and not terminate, the child process. * Remove the _kill_wait() method, replaced with a more agressive close() method. It fixes _make_subprocess_transport() on error. BaseSubprocessTransport.close() calls the close() method of pipe transports, whereas _kill_wait() closed directly pipes of the subprocess.Popen object without unregistering file descriptors from the selector (which caused severe bugs). These changes simplifies the code of subprocess.py.
2015-01-29 19:05:19 -04:00
def _check_proc(self):
if self._proc is None:
raise ProcessLookupError()
2013-10-30 18:56:49 -03:00
def send_signal(self, signal):
asyncio: sync with Tulip Issue #23347: send_signal(), kill() and terminate() methods of BaseSubprocessTransport now check if the transport was closed and if the process exited. Issue #23347: Refactor creation of subprocess transports. Changes on BaseSubprocessTransport: * Add a wait() method to wait until the child process exit * The constructor now accepts an optional waiter parameter. The _post_init() coroutine must not be called explicitly anymore. It makes subprocess transports closer to other transports, and it gives more freedom if we want later to change completly how subprocess transports are created. * close() now kills the process instead of kindly terminate it: the child process may ignore SIGTERM and continue to run. Call explicitly terminate() and wait() if you want to kindly terminate the child process. * close() now logs a warning in debug mode if the process is still running and needs to be killed * _make_subprocess_transport() is now fully asynchronous again: if the creation of the transport failed, wait asynchronously for the process eixt. Before the wait was synchronous. This change requires close() to *kill*, and not terminate, the child process. * Remove the _kill_wait() method, replaced with a more agressive close() method. It fixes _make_subprocess_transport() on error. BaseSubprocessTransport.close() calls the close() method of pipe transports, whereas _kill_wait() closed directly pipes of the subprocess.Popen object without unregistering file descriptors from the selector (which caused severe bugs). These changes simplifies the code of subprocess.py.
2015-01-29 19:05:19 -04:00
self._check_proc()
2013-10-30 18:56:49 -03:00
self._proc.send_signal(signal)
def terminate(self):
asyncio: sync with Tulip Issue #23347: send_signal(), kill() and terminate() methods of BaseSubprocessTransport now check if the transport was closed and if the process exited. Issue #23347: Refactor creation of subprocess transports. Changes on BaseSubprocessTransport: * Add a wait() method to wait until the child process exit * The constructor now accepts an optional waiter parameter. The _post_init() coroutine must not be called explicitly anymore. It makes subprocess transports closer to other transports, and it gives more freedom if we want later to change completly how subprocess transports are created. * close() now kills the process instead of kindly terminate it: the child process may ignore SIGTERM and continue to run. Call explicitly terminate() and wait() if you want to kindly terminate the child process. * close() now logs a warning in debug mode if the process is still running and needs to be killed * _make_subprocess_transport() is now fully asynchronous again: if the creation of the transport failed, wait asynchronously for the process eixt. Before the wait was synchronous. This change requires close() to *kill*, and not terminate, the child process. * Remove the _kill_wait() method, replaced with a more agressive close() method. It fixes _make_subprocess_transport() on error. BaseSubprocessTransport.close() calls the close() method of pipe transports, whereas _kill_wait() closed directly pipes of the subprocess.Popen object without unregistering file descriptors from the selector (which caused severe bugs). These changes simplifies the code of subprocess.py.
2015-01-29 19:05:19 -04:00
self._check_proc()
2013-10-30 18:56:49 -03:00
self._proc.terminate()
def kill(self):
asyncio: sync with Tulip Issue #23347: send_signal(), kill() and terminate() methods of BaseSubprocessTransport now check if the transport was closed and if the process exited. Issue #23347: Refactor creation of subprocess transports. Changes on BaseSubprocessTransport: * Add a wait() method to wait until the child process exit * The constructor now accepts an optional waiter parameter. The _post_init() coroutine must not be called explicitly anymore. It makes subprocess transports closer to other transports, and it gives more freedom if we want later to change completly how subprocess transports are created. * close() now kills the process instead of kindly terminate it: the child process may ignore SIGTERM and continue to run. Call explicitly terminate() and wait() if you want to kindly terminate the child process. * close() now logs a warning in debug mode if the process is still running and needs to be killed * _make_subprocess_transport() is now fully asynchronous again: if the creation of the transport failed, wait asynchronously for the process eixt. Before the wait was synchronous. This change requires close() to *kill*, and not terminate, the child process. * Remove the _kill_wait() method, replaced with a more agressive close() method. It fixes _make_subprocess_transport() on error. BaseSubprocessTransport.close() calls the close() method of pipe transports, whereas _kill_wait() closed directly pipes of the subprocess.Popen object without unregistering file descriptors from the selector (which caused severe bugs). These changes simplifies the code of subprocess.py.
2015-01-29 19:05:19 -04:00
self._check_proc()
2013-10-30 18:56:49 -03:00
self._proc.kill()
async def _connect_pipes(self, waiter):
try:
proc = self._proc
loop = self._loop
asyncio: sync with Tulip Issue #23347: send_signal(), kill() and terminate() methods of BaseSubprocessTransport now check if the transport was closed and if the process exited. Issue #23347: Refactor creation of subprocess transports. Changes on BaseSubprocessTransport: * Add a wait() method to wait until the child process exit * The constructor now accepts an optional waiter parameter. The _post_init() coroutine must not be called explicitly anymore. It makes subprocess transports closer to other transports, and it gives more freedom if we want later to change completly how subprocess transports are created. * close() now kills the process instead of kindly terminate it: the child process may ignore SIGTERM and continue to run. Call explicitly terminate() and wait() if you want to kindly terminate the child process. * close() now logs a warning in debug mode if the process is still running and needs to be killed * _make_subprocess_transport() is now fully asynchronous again: if the creation of the transport failed, wait asynchronously for the process eixt. Before the wait was synchronous. This change requires close() to *kill*, and not terminate, the child process. * Remove the _kill_wait() method, replaced with a more agressive close() method. It fixes _make_subprocess_transport() on error. BaseSubprocessTransport.close() calls the close() method of pipe transports, whereas _kill_wait() closed directly pipes of the subprocess.Popen object without unregistering file descriptors from the selector (which caused severe bugs). These changes simplifies the code of subprocess.py.
2015-01-29 19:05:19 -04:00
if proc.stdin is not None:
_, pipe = await loop.connect_write_pipe(
lambda: WriteSubprocessPipeProto(self, 0),
proc.stdin)
self._pipes[0] = pipe
asyncio: sync with Tulip Issue #23347: send_signal(), kill() and terminate() methods of BaseSubprocessTransport now check if the transport was closed and if the process exited. Issue #23347: Refactor creation of subprocess transports. Changes on BaseSubprocessTransport: * Add a wait() method to wait until the child process exit * The constructor now accepts an optional waiter parameter. The _post_init() coroutine must not be called explicitly anymore. It makes subprocess transports closer to other transports, and it gives more freedom if we want later to change completly how subprocess transports are created. * close() now kills the process instead of kindly terminate it: the child process may ignore SIGTERM and continue to run. Call explicitly terminate() and wait() if you want to kindly terminate the child process. * close() now logs a warning in debug mode if the process is still running and needs to be killed * _make_subprocess_transport() is now fully asynchronous again: if the creation of the transport failed, wait asynchronously for the process eixt. Before the wait was synchronous. This change requires close() to *kill*, and not terminate, the child process. * Remove the _kill_wait() method, replaced with a more agressive close() method. It fixes _make_subprocess_transport() on error. BaseSubprocessTransport.close() calls the close() method of pipe transports, whereas _kill_wait() closed directly pipes of the subprocess.Popen object without unregistering file descriptors from the selector (which caused severe bugs). These changes simplifies the code of subprocess.py.
2015-01-29 19:05:19 -04:00
if proc.stdout is not None:
_, pipe = await loop.connect_read_pipe(
lambda: ReadSubprocessPipeProto(self, 1),
proc.stdout)
self._pipes[1] = pipe
asyncio: sync with Tulip Issue #23347: send_signal(), kill() and terminate() methods of BaseSubprocessTransport now check if the transport was closed and if the process exited. Issue #23347: Refactor creation of subprocess transports. Changes on BaseSubprocessTransport: * Add a wait() method to wait until the child process exit * The constructor now accepts an optional waiter parameter. The _post_init() coroutine must not be called explicitly anymore. It makes subprocess transports closer to other transports, and it gives more freedom if we want later to change completly how subprocess transports are created. * close() now kills the process instead of kindly terminate it: the child process may ignore SIGTERM and continue to run. Call explicitly terminate() and wait() if you want to kindly terminate the child process. * close() now logs a warning in debug mode if the process is still running and needs to be killed * _make_subprocess_transport() is now fully asynchronous again: if the creation of the transport failed, wait asynchronously for the process eixt. Before the wait was synchronous. This change requires close() to *kill*, and not terminate, the child process. * Remove the _kill_wait() method, replaced with a more agressive close() method. It fixes _make_subprocess_transport() on error. BaseSubprocessTransport.close() calls the close() method of pipe transports, whereas _kill_wait() closed directly pipes of the subprocess.Popen object without unregistering file descriptors from the selector (which caused severe bugs). These changes simplifies the code of subprocess.py.
2015-01-29 19:05:19 -04:00
if proc.stderr is not None:
_, pipe = await loop.connect_read_pipe(
lambda: ReadSubprocessPipeProto(self, 2),
proc.stderr)
self._pipes[2] = pipe
assert self._pending_calls is not None
asyncio: sync with Tulip Issue #23347: send_signal(), kill() and terminate() methods of BaseSubprocessTransport now check if the transport was closed and if the process exited. Issue #23347: Refactor creation of subprocess transports. Changes on BaseSubprocessTransport: * Add a wait() method to wait until the child process exit * The constructor now accepts an optional waiter parameter. The _post_init() coroutine must not be called explicitly anymore. It makes subprocess transports closer to other transports, and it gives more freedom if we want later to change completly how subprocess transports are created. * close() now kills the process instead of kindly terminate it: the child process may ignore SIGTERM and continue to run. Call explicitly terminate() and wait() if you want to kindly terminate the child process. * close() now logs a warning in debug mode if the process is still running and needs to be killed * _make_subprocess_transport() is now fully asynchronous again: if the creation of the transport failed, wait asynchronously for the process eixt. Before the wait was synchronous. This change requires close() to *kill*, and not terminate, the child process. * Remove the _kill_wait() method, replaced with a more agressive close() method. It fixes _make_subprocess_transport() on error. BaseSubprocessTransport.close() calls the close() method of pipe transports, whereas _kill_wait() closed directly pipes of the subprocess.Popen object without unregistering file descriptors from the selector (which caused severe bugs). These changes simplifies the code of subprocess.py.
2015-01-29 19:05:19 -04:00
loop.call_soon(self._protocol.connection_made, self)
for callback, data in self._pending_calls:
asyncio: sync with Tulip Issue #23347: send_signal(), kill() and terminate() methods of BaseSubprocessTransport now check if the transport was closed and if the process exited. Issue #23347: Refactor creation of subprocess transports. Changes on BaseSubprocessTransport: * Add a wait() method to wait until the child process exit * The constructor now accepts an optional waiter parameter. The _post_init() coroutine must not be called explicitly anymore. It makes subprocess transports closer to other transports, and it gives more freedom if we want later to change completly how subprocess transports are created. * close() now kills the process instead of kindly terminate it: the child process may ignore SIGTERM and continue to run. Call explicitly terminate() and wait() if you want to kindly terminate the child process. * close() now logs a warning in debug mode if the process is still running and needs to be killed * _make_subprocess_transport() is now fully asynchronous again: if the creation of the transport failed, wait asynchronously for the process eixt. Before the wait was synchronous. This change requires close() to *kill*, and not terminate, the child process. * Remove the _kill_wait() method, replaced with a more agressive close() method. It fixes _make_subprocess_transport() on error. BaseSubprocessTransport.close() calls the close() method of pipe transports, whereas _kill_wait() closed directly pipes of the subprocess.Popen object without unregistering file descriptors from the selector (which caused severe bugs). These changes simplifies the code of subprocess.py.
2015-01-29 19:05:19 -04:00
loop.call_soon(callback, *data)
self._pending_calls = None
asyncio: sync with Tulip Issue #23347: send_signal(), kill() and terminate() methods of BaseSubprocessTransport now check if the transport was closed and if the process exited. Issue #23347: Refactor creation of subprocess transports. Changes on BaseSubprocessTransport: * Add a wait() method to wait until the child process exit * The constructor now accepts an optional waiter parameter. The _post_init() coroutine must not be called explicitly anymore. It makes subprocess transports closer to other transports, and it gives more freedom if we want later to change completly how subprocess transports are created. * close() now kills the process instead of kindly terminate it: the child process may ignore SIGTERM and continue to run. Call explicitly terminate() and wait() if you want to kindly terminate the child process. * close() now logs a warning in debug mode if the process is still running and needs to be killed * _make_subprocess_transport() is now fully asynchronous again: if the creation of the transport failed, wait asynchronously for the process eixt. Before the wait was synchronous. This change requires close() to *kill*, and not terminate, the child process. * Remove the _kill_wait() method, replaced with a more agressive close() method. It fixes _make_subprocess_transport() on error. BaseSubprocessTransport.close() calls the close() method of pipe transports, whereas _kill_wait() closed directly pipes of the subprocess.Popen object without unregistering file descriptors from the selector (which caused severe bugs). These changes simplifies the code of subprocess.py.
2015-01-29 19:05:19 -04:00
except Exception as exc:
if waiter is not None and not waiter.cancelled():
waiter.set_exception(exc)
else:
if waiter is not None and not waiter.cancelled():
waiter.set_result(None)
2013-10-30 18:56:49 -03:00
def _call(self, cb, *data):
if self._pending_calls is not None:
self._pending_calls.append((cb, data))
else:
self._loop.call_soon(cb, *data)
def _pipe_connection_lost(self, fd, exc):
self._call(self._protocol.pipe_connection_lost, fd, exc)
self._try_finish()
def _pipe_data_received(self, fd, data):
self._call(self._protocol.pipe_data_received, fd, data)
def _process_exited(self, returncode):
assert returncode is not None, returncode
assert self._returncode is None, self._returncode
if self._loop.get_debug():
logger.info('%r exited with return code %r', self, returncode)
2013-10-30 18:56:49 -03:00
self._returncode = returncode
if self._proc.returncode is None:
# asyncio uses a child watcher: copy the status into the Popen
# object. On Python 3.6, it is required to avoid a ResourceWarning.
self._proc.returncode = returncode
2013-10-30 18:56:49 -03:00
self._call(self._protocol.process_exited)
self._try_finish()
asyncio: sync with Tulip Issue #23347: send_signal(), kill() and terminate() methods of BaseSubprocessTransport now check if the transport was closed and if the process exited. Issue #23347: Refactor creation of subprocess transports. Changes on BaseSubprocessTransport: * Add a wait() method to wait until the child process exit * The constructor now accepts an optional waiter parameter. The _post_init() coroutine must not be called explicitly anymore. It makes subprocess transports closer to other transports, and it gives more freedom if we want later to change completly how subprocess transports are created. * close() now kills the process instead of kindly terminate it: the child process may ignore SIGTERM and continue to run. Call explicitly terminate() and wait() if you want to kindly terminate the child process. * close() now logs a warning in debug mode if the process is still running and needs to be killed * _make_subprocess_transport() is now fully asynchronous again: if the creation of the transport failed, wait asynchronously for the process eixt. Before the wait was synchronous. This change requires close() to *kill*, and not terminate, the child process. * Remove the _kill_wait() method, replaced with a more agressive close() method. It fixes _make_subprocess_transport() on error. BaseSubprocessTransport.close() calls the close() method of pipe transports, whereas _kill_wait() closed directly pipes of the subprocess.Popen object without unregistering file descriptors from the selector (which caused severe bugs). These changes simplifies the code of subprocess.py.
2015-01-29 19:05:19 -04:00
# wake up futures waiting for wait()
for waiter in self._exit_waiters:
if not waiter.cancelled():
waiter.set_result(returncode)
self._exit_waiters = None
async def _wait(self):
asyncio: sync with Tulip Issue #23347: send_signal(), kill() and terminate() methods of BaseSubprocessTransport now check if the transport was closed and if the process exited. Issue #23347: Refactor creation of subprocess transports. Changes on BaseSubprocessTransport: * Add a wait() method to wait until the child process exit * The constructor now accepts an optional waiter parameter. The _post_init() coroutine must not be called explicitly anymore. It makes subprocess transports closer to other transports, and it gives more freedom if we want later to change completly how subprocess transports are created. * close() now kills the process instead of kindly terminate it: the child process may ignore SIGTERM and continue to run. Call explicitly terminate() and wait() if you want to kindly terminate the child process. * close() now logs a warning in debug mode if the process is still running and needs to be killed * _make_subprocess_transport() is now fully asynchronous again: if the creation of the transport failed, wait asynchronously for the process eixt. Before the wait was synchronous. This change requires close() to *kill*, and not terminate, the child process. * Remove the _kill_wait() method, replaced with a more agressive close() method. It fixes _make_subprocess_transport() on error. BaseSubprocessTransport.close() calls the close() method of pipe transports, whereas _kill_wait() closed directly pipes of the subprocess.Popen object without unregistering file descriptors from the selector (which caused severe bugs). These changes simplifies the code of subprocess.py.
2015-01-29 19:05:19 -04:00
"""Wait until the process exit and return the process return code.
This method is a coroutine."""
if self._returncode is not None:
return self._returncode
waiter = self._loop.create_future()
asyncio: sync with Tulip Issue #23347: send_signal(), kill() and terminate() methods of BaseSubprocessTransport now check if the transport was closed and if the process exited. Issue #23347: Refactor creation of subprocess transports. Changes on BaseSubprocessTransport: * Add a wait() method to wait until the child process exit * The constructor now accepts an optional waiter parameter. The _post_init() coroutine must not be called explicitly anymore. It makes subprocess transports closer to other transports, and it gives more freedom if we want later to change completly how subprocess transports are created. * close() now kills the process instead of kindly terminate it: the child process may ignore SIGTERM and continue to run. Call explicitly terminate() and wait() if you want to kindly terminate the child process. * close() now logs a warning in debug mode if the process is still running and needs to be killed * _make_subprocess_transport() is now fully asynchronous again: if the creation of the transport failed, wait asynchronously for the process eixt. Before the wait was synchronous. This change requires close() to *kill*, and not terminate, the child process. * Remove the _kill_wait() method, replaced with a more agressive close() method. It fixes _make_subprocess_transport() on error. BaseSubprocessTransport.close() calls the close() method of pipe transports, whereas _kill_wait() closed directly pipes of the subprocess.Popen object without unregistering file descriptors from the selector (which caused severe bugs). These changes simplifies the code of subprocess.py.
2015-01-29 19:05:19 -04:00
self._exit_waiters.append(waiter)
return await waiter
asyncio: sync with Tulip Issue #23347: send_signal(), kill() and terminate() methods of BaseSubprocessTransport now check if the transport was closed and if the process exited. Issue #23347: Refactor creation of subprocess transports. Changes on BaseSubprocessTransport: * Add a wait() method to wait until the child process exit * The constructor now accepts an optional waiter parameter. The _post_init() coroutine must not be called explicitly anymore. It makes subprocess transports closer to other transports, and it gives more freedom if we want later to change completly how subprocess transports are created. * close() now kills the process instead of kindly terminate it: the child process may ignore SIGTERM and continue to run. Call explicitly terminate() and wait() if you want to kindly terminate the child process. * close() now logs a warning in debug mode if the process is still running and needs to be killed * _make_subprocess_transport() is now fully asynchronous again: if the creation of the transport failed, wait asynchronously for the process eixt. Before the wait was synchronous. This change requires close() to *kill*, and not terminate, the child process. * Remove the _kill_wait() method, replaced with a more agressive close() method. It fixes _make_subprocess_transport() on error. BaseSubprocessTransport.close() calls the close() method of pipe transports, whereas _kill_wait() closed directly pipes of the subprocess.Popen object without unregistering file descriptors from the selector (which caused severe bugs). These changes simplifies the code of subprocess.py.
2015-01-29 19:05:19 -04:00
2013-10-30 18:56:49 -03:00
def _try_finish(self):
assert not self._finished
if self._returncode is None:
return
if all(p is not None and p.disconnected
for p in self._pipes.values()):
self._finished = True
self._call(self._call_connection_lost, None)
2013-10-30 18:56:49 -03:00
def _call_connection_lost(self, exc):
try:
self._protocol.connection_lost(exc)
finally:
asyncio: sync with Tulip Issue #23347: send_signal(), kill() and terminate() methods of BaseSubprocessTransport now check if the transport was closed and if the process exited. Issue #23347: Refactor creation of subprocess transports. Changes on BaseSubprocessTransport: * Add a wait() method to wait until the child process exit * The constructor now accepts an optional waiter parameter. The _post_init() coroutine must not be called explicitly anymore. It makes subprocess transports closer to other transports, and it gives more freedom if we want later to change completly how subprocess transports are created. * close() now kills the process instead of kindly terminate it: the child process may ignore SIGTERM and continue to run. Call explicitly terminate() and wait() if you want to kindly terminate the child process. * close() now logs a warning in debug mode if the process is still running and needs to be killed * _make_subprocess_transport() is now fully asynchronous again: if the creation of the transport failed, wait asynchronously for the process eixt. Before the wait was synchronous. This change requires close() to *kill*, and not terminate, the child process. * Remove the _kill_wait() method, replaced with a more agressive close() method. It fixes _make_subprocess_transport() on error. BaseSubprocessTransport.close() calls the close() method of pipe transports, whereas _kill_wait() closed directly pipes of the subprocess.Popen object without unregistering file descriptors from the selector (which caused severe bugs). These changes simplifies the code of subprocess.py.
2015-01-29 19:05:19 -04:00
self._loop = None
2013-10-30 18:56:49 -03:00
self._proc = None
self._protocol = None
class WriteSubprocessPipeProto(protocols.BaseProtocol):
def __init__(self, proc, fd):
self.proc = proc
self.fd = fd
2014-01-29 18:22:56 -04:00
self.pipe = None
2013-10-30 18:56:49 -03:00
self.disconnected = False
def connection_made(self, transport):
self.pipe = transport
def __repr__(self):
return f'<{self.__class__.__name__} fd={self.fd} pipe={self.pipe!r}>'
2013-10-30 18:56:49 -03:00
def connection_lost(self, exc):
self.disconnected = True
self.proc._pipe_connection_lost(self.fd, exc)
self.proc = None
2013-10-30 18:56:49 -03:00
def pause_writing(self):
self.proc._protocol.pause_writing()
def resume_writing(self):
self.proc._protocol.resume_writing()
2013-10-30 18:56:49 -03:00
class ReadSubprocessPipeProto(WriteSubprocessPipeProto,
protocols.Protocol):
def data_received(self, data):
self.proc._pipe_data_received(self.fd, data)